Closed rockywind closed 3 years ago
That is expected. Each time the model is trained the results will differ slightly. The model I trained for this code release is slightly different from the model I reported for the paper, as I had to do a refactor for this release that made my old model incompatible.
Yes,thank you very much!
@rockywind hi, may I ask what is your training config? 😊
Thanks for your contribution! The results I tested on the validation set is bellow:
but, the results provided in this paper is:![image](https://user-images.githubusercontent.com/14838781/120129578-d2f02080-c1f6-11eb-91de-22987d902f76.png)